Accumulating evidence reveals a significant correlation between angiopoietin 2 (Ang2) expression and tumor invasion and metastasis in various human cancers, but the major focus of recent studies has been on the angiogenic effects of Ang2. We recently reported that Ang2-stimulated glioma cell invasion results from the up-regulation and activation of matrix metalloprotease 2 (MMP-2) in tumor cells. In this study, we identify a novel mechanism by which Ang2 stimulates MMP-2 expression leading to glioma cell invasion. We show that Ang2 interacts with αvβ1 integrin in Tie2-deficient human glioma cells, activating focal adhesion kinase (FAK), p130Cas, extracellular signal–regulated protein kinase (ERK) 1/2, and c-jun NH2-terminal kinase (JNK) and substantially enhancing MMP-2 expression and secretion. The Ang2/αvβ1 integrin signaling pathway was attenuated by functional inhibition of β1 and αv integrins, FAK, p130Cas, ERK1/2, and JNK. Furthermore, expression of a negative regulator of FAK, FAK-related nonkinase, by U87MG/Ang2–expressing glioma xenografts suppressed Ang2-induced MMP-2 expression and glioma cell infiltration in the murine brain. These data establish a functional link between Ang2 interaction with αvβ1 integrin and glioma cell invasion through the FAK/p130Cas/ERK1/2 and JNK-mediated signaling pathway. (Cancer Res 2006; 66(2): 775-83)
Angiopoietin 2 (Ang2), a known angiogenic regulator, plays important roles in angiogenesis and tumor progression (1). Ang2, expressed in both endothelial cells and tumor cells, has been significantly correlated with tumor metastasis and invasion (2–6), but the mechanism by which Ang2 acts on tumor cells remains obscure. The highly conserved COOH-terminal fibrinogen-like domain of the angiopoietins implies a functional association with the integrin receptor family (7, 8). Integrins are crucial for migration and invasion of tumor cells, not only for mediating adhesion of tumor cells to the extracellular matrix but also for regulating the processes of cell migration and invasion (9). Integrins are heterodimeric glycoproteins composed of one α chain and one β chain. The β1 family of integrins associates with at least one of the 12 α-chains (α1-α11 and αv) representing the major group of cellular extracellular matrix receptors. β1-Integrin promotes metastasis of β1-null tumor cells (9). Mutation of the intracellular domain of β1 integrin differentially affects cell adhesion, invasion, and metastasis (10). Ang2, as well as other members of the angiopoietin family, is a potential substrate for the integrin receptor family in endothelial cells, fibroblasts, and myocytes (11–13). Ang2 not only enhances cell adhesion in both endothelial cells and Tie2-deficient fibroblasts but also triggers integrin-mediated intracellular signal transduction pathways in these cells.
Malignant human gliomas are pathophysiologically characterized by their insidious infiltration of the brain (14). Invasion of glioma cells into adjacent brain structures occurs through the activation of multigenic programs, including matrix metalloproteases (MMP) such as MMP-2, which degrade extracellular matrix to overcome the extracellular matrix barrier at the invasive fronts of tumors (15). Several reports have recently shown that up-regulation of MMP-2 correlates with the invasiveness of human gliomas (16–18). Interestingly, integrin-mediated signaling pathways are involved in regulation of MMP-2 expression and cell invasion in tumor cells (19). Expression patterns of several integrins are correlated with invasive and migratory behaviors in glioma cells in vitro and in vivo (17). MMP-2 expression and tumor cell invasion have also been associated with the activation of downstream regulators of integrins such as focal adhesion kinase (FAK), p130Cas, extracellular signal–regulated protein kinase (ERK)/mitogen-activated protein kinase, and c-jun NH2-terminal kinase (JNK)/stress-activated protein kinase (20, 21). We recently reported that coexpression of Ang2 with MMP-2 was found in the invasive areas, but not in the central regions, of primary human glioma specimens (16, 18). Stable expression of Ang2 by glioma cells leads to an increase in glioma cell invasiveness in vitro as well as in murine intracranial xenografts accomplished through up-regulation and activation of MMP-2 (16).
In this study, we characterize the mechanism by which Ang2 induces glioma cell invasion by stimulating MMP-2 expression and secretion in glioma cells. We show that Ang2 interacts with αvβ1 integrin in U87MG glioma cells, resulting in the activation of FAK, p130Cas, ERK1/2, and JNK, thereby enhancing MMP-2 expression and secretion. As a functional consequence of increased MMP-2 secretion, glioma cell invasion is promoted. Furthermore, stable expression of dominant-negative forms of FAK or p130Cas in glioma cells inhibits Ang2-stimulated MMP-2 expression and secretion by blocking activation of JNK and ERK1/2. Our data suggest that elevation of Ang2 expression in the gliomas leads to increased MMP-2 secretion along with increased cell invasion via αvβ1 integrin/FAK–mediated activation of ERK1/2 and JNK.
Materials and Methods
Antibodies and reagents. Human U87MG and T98G glioma cells were obtained from American Type Culture Collection (Manassas, VA), U251MG and U373MG glioma cells were from our collection, and their culture was previously described (16). The following reagents were used for this study: anti-Ang2 (C-19), anti-p130Cas (C-20), anti-JNK1 (C-17), anti-FAK (C-20), and anti-β3 integrin (H-96) antibodies (Santa Cruz Biotechnology, Santa Cruz, CA); neutralizing anti-β3 integrin (B3A), anti-β1 integrin (JB1A), and anti-α3 integrin antibodies (P1B5); a β1 integrin polyclonal investigator kit; an α integrin blocking investigator kit that includes anti-β1, anti-α1, anti-α2, anti-α3, anti-α4, anti-α5, anti-α6, anti-αv, anti-αvβ3, and anti-αvβ5 antibodies; FITC-conjugated anti-αvβ3 (clone LM609) and anti-αvβ5 antibodies (clone P1F6; both for fluorescence-activated cell sorting analyses) and anti-MMP-2 antibodies (Chemicon International, Inc., Temecula, CA); FITC-conjugated anti-β1 antibody (clone 4B4-FITC, for FACS analyses, Coulter Corp., Miami, FL); anti-Ang2 antibody and human recombinant Ang2 protein (R&D Systems, Minneapolis, MN); phosphospecific antibodies against activated ERK1/2 (E10 to pT202/pY204) and activated JNK (G9 to pT183/pY185, Cell Signaling, Beverly, MA); anti-phosphotyrosine (pY) antibody (4G10, Upstate Biotechnology, Lake Placid, NY); anti-FAK (pY397 and pY861) antibodies (Biosource, Camarillo, CA); anti-hemagglutinin (HA) antibody (12CA5, Covance, Berkeley, CA) and anti–Myc tag antibody (Medical & Biological Laboratory International, Woburn, MA). Fibronectin, RGD, RAD, U0126, SP600125, and MMPi-III were from Calbiochem (San Diego, CA). Protein G-plus and Protein G-plus/Protein A agarose bead slurry were from Oncogene Research (San Diego, CA). Ni+-NTA beads were from Qiagen (Valencia, CA). Alexa Fluor 594–conjugated goat anti-mouse immunoglobulin G (IgG) and Alexa Fluor 594–conjugated donkey anti-goat IgG antibodies were from Molecular Probes, Inc. (Carlsbad, CA). Other reagents were from Invitrogen (Carlsbad, CA), Sigma (St. Louis, MO), Fisher Scientific (Hampton, NH), or BD Biosciences (San Diego, CA).
Cell adhesion assay. Microtiter plates were coated with 10 μg/mL Ang2, fibronectin, or bovine serum albumin (BSA). Cells were harvested, preincubated in serum-free DMEM with or without various treatments for 30 minutes on ice, seeded into the plate, and incubated for another 30 minutes at 37°C. Nonattached cells were removed by vigorous agitation and aspiration. Attached cells were fixed and stained with crystal violet (0.2% in H2O). Cell-associated crystal violet was eluted by addition of 100 μL of 10% acetic acid. Cell adhesion was quantified by measuring the absorbance of the eluted crystal violet at a wavelength of 600 nm using a microplate reader (22).
Pull-down assay and coimmunoprecipitation. Pull-down assay and coimmunoprecipitation for integrin and ligand interaction were done as previously described (23). Briefly, cultured U87MG cells were incubated with conditioned medium (CM)/Ang2 or CM/control at 37°C/5% CO2 for 2 hours and washed thrice with TBS containing CaCl2 and MnCl2 (1 mmol/L each). In the pull-down assays, the cells were lysed with a buffer containing 20 mmol/L Tris-HCl (pH 8.0), 300 mmol/L NaCl, 1 mmol/L CaCl2, 1 mmol/L MnCl2, 30 mmol/L imidazole, 1.5% Triton X-100, and 2 mmol/L phenylmethylsulfonyl fluoride (PMSF) and centrifuged. The supernatant was rotated overnight with Ni+-NTA beads at 4°C. The incubated beads were extensively washed with a wash buffer containing 25 mmol/L NaHPO4, 150 mmol/L NaCl, 2 mmol/L CaCl, and 45 mmol/L imidazole, eluted with the wash buffer containing 500 mmol/L imidazole, and precipitated with trichloroacetic acid. In coimmunoprecipitation assays, the cells were lysed with a lysis buffer containing 20 mmol/L Tris-HCl (pH 7.4), 150 mmol/L NaCl, 1 mmol/L CaCl2, 2 mmol/L MnCl2, 1.5% Triton X-100, 2 mmol/L PMSF, and 10 μg/mL leupeptin and precleared with various nonspecific IgG and either Protein G-plus or Protein G-plus/Protein A agarose beads. The precleared cell lysate containing 1 mg of total protein was separately incubated overnight at 4°C with 5 μg different antibodies [anti-β1 (JB1A), anti-β3 (B3A), anti-α3, anti-α5, and anti-αv integrin antibodies that were included in the α integrin blocking investigator kit, anti-Myc (9B11), and anti-Ang2 antibodies], and then collected with 50 μL of either a Protein G-plus or a Protein G-plus/Protein A agarose bead slurry. The eluted proteins from Ni+-NTA beads or the antibody-complexed proteins from coimmunoprecipitation were subjected to SDS-PAGE under reducing conditions and detected with an anti-β1 integrin (JB1A) or an anti-Ang2 antibody.
Gelatin zymography and in vitro invasion assays. Zymographic analyses for MMP-2 and in vitro invasion assays were done as previously described (16).
Intracranial brain tumor xenografts and immunohistochemistry. Stereotactic implantation of various types of U87MG cells, sacrifice of glioma-bearing mice, tumor processing, and analyses of intracranial U87MG glioma invasiveness were done as previously described (16).
Ang2-enhanced adhesion of Tie2-deficient glioma cells is mediated by integrins. We first used FACS, immunoblot, and reverse transcription-PCR (RT-PCR) analyses (see Supplementary materials) to assess the expression of 11 integrin subunits and complexes that have been implicated in cancer cell migration and invasion (19). RT-PCR analyses were done to further confirm the absence of the particular integrin subunits in U87MG cells that had shown negative results by immunoblot analysis. We found that β1, β3, α2, α3, α5, αv, α6, αvβ3, and αvβ5 are expressed in U87MG cells whereas the expression of β4, α1, and α4 in U87MG cells was undetectable (Supplementary Table S1). Then, we purified recombinant Ang2 from the CM of U87MG/Ang2–expressing (U87MG/Ang2) cells using Ni+-NTA affinity chromatography, achieving an ∼60-fold increase in Ang2 protein concentration with 90% purity (data not shown). The biological activity of the purified Ang2 was verified by competitive inhibition of Ang1-stimulated tyrosine phosphorylation of Tie2 in endothelial cells (16).
Next, we assessed whether Ang2 stimulates U87MG cell adhesion and which integrin subunit is involved in Ang2-stimulated cell adhesion and motility. As shown in Supplementary Fig. S1, U87MG cells tightly adhered to the Ang2-coated surface compared with the BSA-coated surface. A general integrin blocking reagent, EDTA caused complete inhibition of U87MG cell adherence to Ang2, and a RGD peptide showed more potent inhibitory effect when compared with a RAD peptide. We also assessed whether the integrin receptors expressed in U87MG glioma cells (Supplementary Table S1) are involved in Ang2-stimulated cell adhesion by preincubating the cells with functional blocking integrin antibodies.
Among these functional blocking antibodies, only an anti-β1 antibody caused complete inhibition of U87MG cells adhering to Ang2 compared with mouse IgG control. Furthermore, anti-β3, anti-α2, anti-α3, anti-α5, anti-α6, and anti-αv antibodies blocked Ang2-stimulated U87MG cell adhesion to a lesser degree compared with the anti-β1 antibody. These data suggest that interaction of Ang2 with integrins is involved in Ang2-stimulated U87MG cell adhesion and β1, β3, α3, α5, and αv integrins are likely involved in this adhesion process.
Ang2 associates with β1, α5, and αv integrins in Tie2-deficient human glioma cells. To evaluate whether Ang2 is associated with β1, β3, α3, α5, and αv integrins in Tie2-deficient glioma cells, we did pull-down and coimmunoprecipitation assays that have been used to identify integrin-associated proteins (23). Integrin subunits β1, α5, αv, but not β3, α3, and Ang2, were detected in the pull-down portion by immunoblot analyses of U87MG cells that had been incubated with serum-free CM of U87MG/Ang2–expressing cells (CM/Ang2; Fig. 1A). In contrast, these integrins and Ang2 were undetectable in the pull-down portion from the cells that were incubated with the serum-free CM of U87MG cells (CM/control) whereas these integrins were found in the total U87MG cell lysates without exposure to CM. In the immunoprecipitation assays, equal amounts of β1 integrin proteins in the immunoprecipitated complexes from both groups were seen by immunoblot analysis, but Ang2 was only present in the immunoprecipitated complexes from the U87MG cells incubated with CM/Ang2 (Fig. 1B,, a). In parallel immunoprecipitation experiments using an anti-Myc antibody or an anti-Ang2 antibody, only the immunoprecipitates from U87MG cells exposed to CM/Ang2 reacted strongly with an anti-β1 antibody (Fig. 1B , b and c).
αvβ1 integrin mediates Ang2-stimulated tyrosine phosphorylation of FAK and p130Cas. We analyzed whole-cell lysates by immunoblot analyses with an anti-phosphotyrosine antibody. As shown in Fig. 2A, increased tyrosine phosphorylation (pTyr) of proteins with molecular weights in the range of 105 to 160 kDa and 70 kDa was observed in the U87MG cells adherent to Ang2. Anti-β1 and anti-αv antibodies attenuated the Ang2-enhanced tyrosine phosphorylation of proteins in U87MG cells whereas the anti-β3, anti-α3, and anti-α5 antibodies had no effect on tyrosine phosphorylation. We also found that tyrosine phosphorylation of proteins in Ang2 stably transfected U87MG cells was constitutively up-regulated, which may explain why these cells adhere to cell culture plates in a more robust fashion than parental U87MG cells. These results suggest that the αvβ1 integrin may be the principal integrin receptor that is involved in Ang2-stimulated tyrosine phosphorylation of multiple proteins in U87MG cells.
We then assessed whether Ang2 stimulates tyrosine phosphorylation of FAK and p130Cas through the association with αvβ1 integrin in U87MG cells. As shown in Fig. 2B, the extent of phosphorylation of FAK was enhanced in U87MG cells seeded onto Ang2 as compared with BSA, whereas phosphorylation of p130Cas was dramatically induced in the cells plated onto Ang2. Pretreatment of U87MG cells with neutralizing anti-β1 or anti-αv antibodies inhibited phosphorylation of both proteins enhanced by Ang2 compared with the control (mouse IgG) whereas inhibition of β3 and α5 had little or no effect. These results indicate that Ang2 stimulates tyrosine phosphorylation of FAK and p130Cas through αvβ1 integrin in U87MG cells.
Next, we established U87MG cell clones that stably expressed FAK-related nonkinase (FRNK, an endogenous FAK inhibitor; ref. 24) or its mutant form, FRNK S-1034, in U87MG cells or U87MG/Ang2 cells (Fig. 2C). Expression of FRNK or FRNK S-1034 in U87MG or U87MG/Ang2 cells did not result in any changes in cell morphology, cell growth rate, or expression levels of Ang2 compared with that of parental U87MG or U87MG/Ang2 cells (data not shown). Compared with Ang2-treated U87MG cells, decreased tyrosine phosphorylation of FAK and p130Cas was seen in FRNK-expressing cell clones (F19, Ang2-F1; Fig. 2D). Interestingly, expression of FRNK in U87MG and U87MG/Ang2 cells suppressed Ang2-stimulated tyrosine phosphorylation of FAK at amino acids Tyr397 and Tyr861. As expected, FRNK S-1034 expression (S11 or Ang2-S10) did not affect Ang2-induced tyrosine phosphorylation of Tyr861 of FAK and p130Cas. These results show that adhesion to Ang2 increases tyrosine phosphorylation of FAK and p130Cas in U87MG cells and Ang2-induced tyrosine phosphorylation of p130Cas is mediated through FAK activation.
Ang2 activates JNK1 and ERK1/2 through the αvβ1 integrin/FAK/p130Cas signaling pathway. We examined the activation of JNK1 and ERK1/2 in U87MG cells adhering to Ang2. As shown in Fig. 3A, increased phosphorylation of both JNK1 and ERK1/2 in U87MG cells was observed as early as 20 minutes and peaked between 40 and 60 minutes after the cells adhered to Ang2. The Ang2-stimulated activation of JNK1 and ERK1/2 in U87MG cells was mediated by αvβ1 integrin because inhibition of αv or β1, but not α5 or β3, attenuated the phosphorylation of both JNK1 and ERK1/2 compared with control mouse IgG (Fig. 3C and D). The Ang2 stimulation of JNK1 and ERK1/2 in U87MG cells was also specific because a JNK inhibitor, SP600125 (Fig. 3C; ref. 25), and an ERK1/2 inhibitor, U0126 (Fig. 3D; refs. 26, 27), completely blocked the protein phosphorylation induced by Ang2 compared with the controls.
Next, we determined whether p130Cas is an important mediator of the Ang2/αvβ1/FAK signaling pathway by expressing the Src homology 3 (SH3) domain of p130Cas (CasSH3) in U87MG cells. CasSH3 binds to FAK and inhibits endogenous p130Cas binding, thus acting in a dominant-negative manner (28). Expression of CasSH3 in U87MG cells suppressed Ang2-stimulated tyrosine phosphorylation of p130Cas (Fig. 3B) but did not cause any morphologic changes or alteration in cell growth rate compared with U87MG parental cells (data not shown).
Because JNK1 and ERK1/2 are potential downstream effectors of FAK and p130Cas in mediating cell motility and invasion (21), we tested whether expression of CasSH3 or FRNK in U87MG cells blocks Ang2-stimulated activation of JNK1 and ERK1/2. Expression of FRNK (F19) or CasSH3 (C8) in U87MG cells inhibited Ang2-stimulated activation of both JNK1 (Fig. 3C) and ERK1/2 (Fig. 3D) compared with controls (mouse IgG, DMSO, and a FRNK S-1034–expressing cell clone, S11). In parallel, expression of FRNK in U87MG/Ang2 cell clones (Ang2-F1 and Ang2-F7) also attenuated Ang2-stimulated activation of JNK1 and ERK1/2 when compared with parental U87MG/Ang2 cells or U87MG/Ang2/FRNK S-1034 cell clones, Ang2-S2 and Ang2-S10 (data not shown). These data suggest that Ang2 activates JNK1 and ERK1/2 in U87MG cells and the activation is mediated by the Ang2/αvβ1 integrin/FAK/p130Cas signaling pathway.
Ang2 stimulates MMP-2 expression and secretion via αvβ1 integrin/FAK/p130Cas–mediated activation of ERK1/2 and JNK1. We examined whether stimulation of MMP-2 by Ang2 is mediated by the αvβ1 integrin/FAK/p130Cas signaling pathway through activation of JNK1 and ERK1/2 in glioma cells. As shown in Fig. 4A, an increased amount of pro-MMP-2 (72 kDa) was detected in the serum-free CM of U87MG cells plated onto Ang2-coated plates by gelatin zymography and immunoblot analyses in a concentration-dependent manner compared with the cells seeded onto BSA-coated plates. The active form of MMP-2 (64 kDa) was almost undetectable in the CM of U87MG cells plated either on Ang2 or BSA, probably due to insufficient MMP-2 activation resulting from a short period of cell culture (18 hours). Inhibition of β1 or αv, but not α5, by neutralizing antibodies significantly suppressed Ang2-stimulated MMP-2 secretion (Fig. 4B). Both SP600125 and U0126 also effectively inhibited Ang2-stimulated MMP-2 secretion in U87MG cells (Fig. 4B).
We next assessed whether inhibition of FAK and p130Cas by FRNK and CasSH3 blocks Ang2-stimulated MMP-2 expression and secretion (Fig. 4C). We found that MMP-2 expression and secretion were decreased in U87MG/FRNK–expressing cell clones (F13 and F19) and U87MG/CasSH3–expressing cell clones (C3 and C8) compared with parental U87MG cells or U87MG/FRNK S-1034–expressing cell clones (S11 and S21) exposed to Ang2. Lastly, we determined whether Ang2 stimulates MMP-2 secretion via β1 integrin in other Tie2-deficient human glioma cell lines. As shown in Fig. 4D, preincubation of T98G, U251MG, and U373MG glioma cells with an anti-β1 antibody significantly inhibited Ang2-stimulated MMP-2 secretion. Additionally, expression of MMP-2 in U87MG cells exposed to various treatments was also assessed with immunoblot analyses using an anti-MMP-2 antibody. The detected levels of expression of MMP-2 were correlated to the amounts of secretion of MMP-2 in the CM by the gelatin zymographic analyses (Fig. 4A-D). Taken together, our results show that Ang2 is capable of stimulating the expression and secretion of MMP-2 in Tie2-deficient glioma cells through αvβ1 integrin/FAK/p130Cas–mediated activation of ERK1/2 and JNK1. Interestingly, we could not detect MMP-9 by gelatin zymography from CM of the various cells after an 18-hour cell culture with the various treatments, suggesting that MMP-9 is not involved in Ang2-stimulated human glioma invasion nor regulated by the Ang2/αvβ1 integrin/FAK/p130Cas/ERK1/2 and JNK1 signaling pathway (data not shown).
Ang2 stimulates cell invasion via the αvβ1 integrin/FAK/p130Cas signaling pathway through ERK1/2 and JNK1 activation. To determine whether Ang2 stimulation of the αvβ1 integrin/FAK/p130Cas signaling pathway leads to increased U87MG glioma cell invasion, we did in vitro cell invasion assays. As shown in Fig. 5A, exposure to exogenous Ang2 promoted cell invasion whereas Ang2-stimulated cell invasion was blocked by a MMP inhibitor (MMPi-III, 50 μmol/L) that preferentially attenuates MMP-2 activity (Calbiochem; ref. 16), anti-β1 and anti-αv antibodies (25 μg/mL), SP600125 (25 μmol/L), and U0126 (25 μmol/L) compared with cells treated with mouse IgG or DMSO. Expressing FRNK (F13 and F19) and CasSH3 (C3 and C8) in U87MG cells also suppressed Ang2-stimulated cell invasion compared with parental U87MG cells or U87MG/FRNK S-1034–expressing cells (S11 and S21).
Next, we assessed whether inhibition of β1, αv integrin, FAK, JNK1, ERK1/2, and MMP-2 in U87MG/Ang2–expressing cells (Ang2) suppresses Ang2-induced cell invasion. As shown in Fig. 5B, FRNK expression in U87MG/Ang2 expressing cells (Ang2-F1 and Ang2-F7) effectively suppressed Ang2-stimulated cell invasion compared with U87MG/Ang2–expressing cells (Ang2) and U87MG/Ang2/FRNK S-1034–expressing clones (Ang2-S2 and Ang2-S10). Similar to the results with parental U87MG cells seeded on an Ang2-coated surface described in Fig. 4A, pretreatment of U87MG/Ang2–expressing cells with MMPi-III (50 μmol/L), anti-β1 and anti-αv antibodies (25 μg/mL), SP600125 (25 μmol/L), or U0126 (25 μmol/L) also inhibited the cell invasion enhanced by expressing Ang2 in U87MG cells (Fig. 5B). These data suggest that both the αvβ1 integrin/FAK/p130Cas/ERK1/2 and JNK1 signaling pathway and MMP-2 are involved in Ang2-stimulated glioma cell invasion.
Inhibition of FAK suppresses Ang2-stimulated MMP-2 expression and glioma invasion in vivo. Lastly, a glioma xenograft model in the murine brain was employed to determine whether inhibition of FAK by FRNK expression could inhibit Ang2-stimulated glioma cell infiltration in vivo. As shown in Fig. 6, mice that received U87MG/Ang2-expressing cells developed invasive gliomas (Fig. 6B). The U87MG/Ang2–expressing cells that additionally express FRNK did not exhibit an invasive phenotype and formed well-circumscribed gliomas (Fig. 6C), which are analogous to U87MG/LacZ–expressing tumors (Fig. 6A). In contrast, the expression of FRNK S-1034 in U87MG/Ang2–expressing cells (Ang2-S2 and Ang2-S10) was unable to inhibit Ang2-stimulated invasion (Fig. 6D). Immunohistochemical and immunoblot analyses using anti-Ang2 and anti-MMP-2 antibodies show that expression of FRNK or FRNK S-1034 had no affect on Ang2 expression (Fig. 6G and H) compared with that of U87MG/Ang2-expressing tumors (Fig. 6F). However, FRNK expression did noticeably reduce the MMP-2 expression (Fig. 6K) compared with that of U87MG/Ang2 expressing tumors (Fig. 6J) whereas there was no inhibition of MMP-2 expression by FRNK S-1034 (Fig. 6L). Immunoblot analyses confirmed that expression of FRNK or FRNK S-1034 by these cell clones did not alter expression levels of Ang2 but FRNK expression did inhibit Ang2-stimulated MMP-2 expression compared with that in U87MG/Ang2–expressing cells (data not shown). The sensitivity of detecting MMP-2 expression in established U87MG gliomas by immunohistochemical staining (Fig. 6I, to L) is lower than that in cultured U87MG cells by zymographic and Western blot assays (Fig. 4A-C). Expression of HA-tagged FRNK and FRNK S-1034 by U87MG/Ang2–expressing cells in vivo was detected by immunohistochemistry using an anti-HA tag antibody (Fig. 6M and N). Furthermore, although expression of FRNK, but not FRNK S-1034, in U87MG/Ang2–expressing cells inhibited Ang2-stimulated glioma cell invasion in vitro and in vivo, no alteration of tumor formation of intracranial gliomas formed by either U87MG/Ang2/FRNK–expressing or U87MG/Ang2/FRNK S-1034–expressing cells was observed. Together, these results show that inhibition of FAK by FRNK in U87MG/Ang2–expressing cells suppressed Ang2-stimulated MMP-2 expression and glioma cell infiltration in vivo.
Growth factors and angiopoietin family members recently have emerged as novel ligands for integrins eliciting specific biological effects. For example, latent forms of transforming growth factor β directly bind to integrin αvβ6 (29) and αvβ1 (30) in regulating pulmonary inflammation and fibrosis. VEGF165 and VEGF189 promote endothelial cell adhesion, migration, and survival through interaction with several integrins independent of VEGF receptors (31). Fibrinogen has been shown as a ligand for several integrins including the α5β1 integrin receptor (32). Members of the angiopoietin family contain a fibrinogen-like domain at the COOH terminus of their proteins (8, 33), suggesting integrins as potential candidate receptors (7). Adhesion of Tie2-deficient cells, such as fibroblasts and skeletal myocytes, to Ang2-coated as well Ang1-coated surfaces was recently shown (11, 12). Lacking Tie2 receptors, fibroblasts and myocytes adhere to Ang2- or Ang1-coated surfaces and the adhesion is sensitive to the inhibition of β1, β3, α6, and αv integrins. Interactions of Ang1 and/or Ang2 with these integrins activate FAK and ERK1/2, increasing cell adhesion (11), and Akt, FAK, and ERK1/2, promoting cell survival (12). Even in endothelial cells, Ang1 is able to mediate selectively α5β1 integrin outside-in FAK activation, leading to a cross-talk of Tie2 and α5β1, and fine-tuning modulation of the vascular effect (13). In this study, we report that Ang2 associates with αvβ1 integrin in Tie2-deficient glioma cells. We observed that Ang2 induces extracellular adhesion and activates intracellular signal transduction pathways in glioma cells. Pull-down and immunoprecipitation assays showed that αvβ1 integrin could be copurified with Ang2. The association of Ang2 with αvβ1 integrin was highly calcium and manganese dependent in both pull-down and coimmunoprecipitation assays, suggesting the association of Ang2 with αvβ1 integrin is similar to that of Ang1, Ang2 to various integrins (12), and PG-MV/vesican with β1 integrin (23). Our data further show that Ang2 stimulates αvβ1 integrin/FAK–mediated signaling, leading to an increase in MMP-2 expression and glioma cell invasion.
FAK and p130Cas are involved in early integrin receptor signaling. Up-regulated FAK expression and activation have been found in various types of human cancers. An increase of FAK phosphorylation on Tyr397 and Tyr861 strongly correlates with the acquisition of an invasive phenotype of tumor cells (21, 34). We found that Ang2 stimulates FAK tyrosine phosphorylation on these two tyrosine residues in U87MG cells. Expression of FRNK, but not its mutant form FRNK S-1034, suppressed Ang2-induced activation of FAK/p130Cas/ERK1/2 and JNK1, MMP-2 expression and secretion, and cell invasion in vitro and reversed the phenotype of Ang2-stimulated glioma cell infiltration and MMP-2 expression in vivo. These results are consistent with previous reports that FAK Tyr397 phosphorylation is required to promote JNK activation and cell invasion by v-Src (35) and that FRNK expression suppressed Src-stimulated Try861 phosphorylation, inhibiting Src activation of FAK downstream signaling, MMP-2 expression, and tumor metastasis (25). Moreover, FRNK overexpression disrupts focal adhesion and induces anoikis in ventricular myocytes by inhibiting endothelin-1-induced phosphorylation of proline-rich tyrosine kinase 2 (PyK2; ref. 36). In glioma cells, PyK2 plays critical roles in cell migration. Overexpression of PyK2 stimulates glioma cell migration whereas inhibition of PyK2 by FRNK expression or siRNA for PyK2 significantly inhibits glioma cell migration (37, 38). Therefore, it is plausible that in addition to suppression of FAK, FRNK expression in U87MG/Ang2–expressing cells also inhibits PyK2 activation, leading to decreased Ang2-induced glioma cell invasion in vitro and in vivo.
p130Cas is an adaptor protein that associates with FAK directly through its SH3 domain to a proline-rich region of FAK (amino acids 712-718; ref. 21). p130Cas also interacts with FAK indirectly through binding to Src family kinase via its SH3 domain. Src stimulates FAK Tyr397 phosphorylation, resulting in binding of FAK to an SH2 domain of the Src protein (39). Our data show that Ang2-induced p130Cas tyrosine phosphorylation was completely abolished by neutralizing anti-β1 and anti-αv antibodies or FRNK, but not FRNK S-1034, suggesting that the interaction of phosphorylated FAK at Tyr397 with Src family kinase is also involved in Ang2/αvβ1/FAK/p130Cas signaling (21). Additionally, expression of a dominant-negative regulator for p130Cas, CasSH3 (28), significantly attenuated Ang2-induced phosphorylation of p130Cas and downstream signaling activities in U87MG cells, showing that the interaction between FAK and p130Cas mediates Ang2/αvβ1 integrin stimulation of glioma cell invasion.
Previous studies have shown that integrin-mediated stimulation of FAK/p130Cas signaling pathways activates ERK1/2 and JNK1 (19), MMP-2 secretion, and cell invasion (25, 35, 40, 41). Our results identify an αvβ1 integrin/FAK/p130Cas/ERK1/2 and JNK1 signaling pathway that mediates Ang2-stimulated MMP-2 expression, secretion, and cell invasion. Functional blocking antibodies for Ang2 and αv and β1 integrins inhibited Ang2 stimulation of ERK1/2 and JNK. FRNK and CasSH3 or inhibitors for ERK1/2 (U0126) and JNK1 (SP600125) suppressed Ang2 activation of ERK1/2 and JNK1. Furthermore, Ang2-stimulated MMP-2 expression and cell invasion were attenuated by αv and β1 inhibition, FRNK, CasSH3, U0126, and SP600125. These results are consistent with previous findings that both ERK1/2 and JNK mediate Src-stimulated v-Src3T3 cell invasion and FRNK inhibits v-Src-stimulated phosphorylation of ERK1/2 and JNK (25). Moreover, our results differ from a study showing that only JNK1 activation, and not ERK1/2 phosphorylation, is required in v-Src-stimulated FAK-Src-p130Cas signaling in FAK-null fibroblasts (35).
β1-Integrin has been shown to activate FAK-mediated signaling pathways leading to promotion of invasion of various types of human carcinoma cells (10, 42, 43). Suppression of β1 expression in rat C6 glioma cells prevented diffuse glioma cell invasion in the brain (44). On the other hand, several α integrin subunits that associate with the β1 integrin subunit, such as αv, α5, and α6 integrins, have been implicated in mediating tumor cell invasion (19, 42, 43). In addition, αvβ3 integrin is also associated with invasion of human gliomas (17, 45). In our study, we show that αvβ1, but not other integrin receptors, is associated with Ang2-stimulated glioma cell invasion. Moreover, we cannot rule out other molecules involved in Ang2-stimulated glioma cell invasion. For example, Ang2 may associate with αvβ5 integrin because functional blocking of αv, α5, β1, β3, and αvβ5 inhibited fibroblast or myocyte adhesion to Ang2 (11, 12). ANGPTL3 and tenascin C interact with αvβ3 integrin through the fibrinogen-like domain that is also present in Ang2 (7, 8). We found that α3 and β3 are not associated with Ang2 whereas α5 is weakly associated with Ang2. However, inhibition of α3, α5, and β3 only moderately suppressed Ang2-mediated cell adhesion (Supplementary Table S1 and Fig. S1), but not phosphorylation of FAK and p130Cas (Fig. 2). Thus, αvβ3 and α5β1 integrins are probably not involved in Ang2-stimulated glioma cell invasion in our model. Additionally, we have recently reported that MT1-MMP, laminin-5 γ2, and Ang2 are coexpressed in the invasive regions, but not in the central areas, of primary glioma specimens and exogenous Ang2 stimulates expression and activation of both MT1-MMP and laminin-5 γ2 in U87MG/Ang2–expressing glioma xenografts in the murine brain and in vitro (18). Taken together, these results warrant further investigation to determine other signaling pathways involved in Ang2-induced glioma cell invasion.
In summary, this study identifies a novel molecular mechanism by which Ang2 stimulates human glioma cell invasion via the αvβ1 integrin/FAK/p130Cas/ERK1/2 and JNK1 signaling pathway and MMP-2 expression (Supplementary Fig. S2). Considering the fact that Ang2 expression has significant links to tumor invasion and metastases in various types of human cancers (2–6), our studies provide a basis for further characterization of the molecular cascades involved in Ang2-stimulated glioma invasion, which may yield promising insight for the development of new diagnostic and therapeutic applications for invasive tumors.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
Grant support: James S. McDonnell Foundation grant and NIH grants CA102011 and CA102310 (D.D. Schlaepfer) and an ACS grant RSG CSM107144 (S-Y. Cheng).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Dr. Kari Alitalo for an expression vector containing human Ang2 cDNA and Dr. Jun-Lin Gun for an expression vector containing human CasSH3 cDNA.